Alterations in the Fungal Microbiome in Ulcerative Colitis

Abstract

Background

Although gut fungi have been implicated in the immunopathogenesis of inflammatory bowel disease, the fungal microbiome has not been deeply explored across endohistologic activity and treatment exposure in ulcerative colitis.

Methods

We analyzed data from the SPARC IBD (Study of a Prospective Adult Research Cohort with Inflammatory Bowel Disease) registry. We evaluated the fungal composition of fecal samples from 98 patients with ulcerative colitis across endoscopic activity (n = 43), endohistologic activity (n = 41), and biologic exposure (n = 82). Across all subgroups, we assessed fungal diversity and differential abundance of taxonomic groups.

Results

We identified 500 unique fungal amplicon sequence variants across the cohort of 82 patients, dominated by phylum Ascomycota. Compared with endoscopic remission, patients with endoscopic activity had increased Saccharomyces (log2 fold change = 4.54; adjusted P < 5 × 10^-5) and increased Candida (log2 fold change = 2.56; adjusted P < .03). After adjusting for age, sex, and biologic exposure among patients with endoscopic activity, Saccharomyces (log2 fold change = 7.76; adjusted P < 1 × 10^-15) and Candida (log2 fold change = 7.28; adjusted P< 1 × 10^-8) remained enriched during endoscopic activity compared with quiescence.

Conclusions

Endoscopic inflammation in ulcerative colitis is associated with an expansion of Saccharomyces and Candida compared with remission. The role of these fungal taxa as potential biomarkers and targets for personalized approaches to therapeutics in ulcerative colitis should be evaluated.

Graphical Abstract

Key Messages.

What is already known?

• The fungal microbiome is altered in inflammatory bowel disease compared with healthy controls, with specific taxa, including Saccharomyces and Candida increased in inflammatory bowel disease.

What is new here?

• Saccharomyces and Candida are increased during endoscopic activity in ulcerative colitis compared with remission, even after controlling for the effects of medication exposure.

How can this study help patient care?

• Fungal biomarkers—including levels of Candida and Saccharomyces—may help physicians better predict disease outcomes or personalize therapies for their patients.

Introduction

Gut fungi have been increasingly implicated in the immunopathogenesis of ulcerative colitis (UC). In prior investigations, circulating antibodies directed against Saccharomyces cerevisiae (anti-S. cerevisiae [ASCA]) have been observed in ~10% to 15% of UC patients.^1-3 Genetic polymorphisms in fungal-sensing genes, such as Dectin-1, are associated with severe forms of UC; abrogating Dectin-1 in murine models increases susceptibility to chemical colitis.⁴ Specific fungal organisms, including Saccharomyces and Candida, have been shown to modulate intestinal inflammation and are implicated in the pathogenesis of UC.^5,6

Supporting these observations, direct evaluation of the fungal microbiome (mycobiome) through deep sequencing reveals alterations in fungal communities among UC patients compared with healthy individuals, with observed reductions in fungal diversity.⁷ However, evaluation of the fecal mycobiome in UC remains limited. Most deep sequencing studies have combined patients with UC and Crohn’s disease; few studies have examined changes in the UC mycobiome with respect to validated endoscopic indices (Mayo endoscopic subscore [MES]) or histologic parameters; furthermore, the effects of treatment exposure, including biologic medications, on fungal communities in UC also remains understudied. Additionally, recent investigations suggest that the diversity and constituents of gut fungi may influence success rates of fecal microbiota transplantation (FMT) in UC⁸—with high Candida abundance pre-FMT in the recipient predictive of clinical response and low Candida abundance post-FMT suggestive of ameliorated disease severity.

Our study aimed to map fungal communities among 98 UC patients with varying levels of endoscopic activity, endohistologic activity, and treatment exposure. We hypothesized that Candida would be increased during endoscopic activity. Defining mycobial populations in UC may allow for the characterization of fungal-UC phenotypes that might direct personalized approaches to therapies, including predicting the success of FMT, probiotics, or antimicrobial treatments.

Methods

Cohort Description

Our cohort was derived from the SPARC IBD (Study of a Prospective Adult Research Cohort with IBD) registry, a geographically diverse longitudinal research cohort utilizing standardized data and biosample collection methods and processing techniques.⁹ The primary objective of the SPARC IBD registry has been to identify clinical and molecular biomarkers to better inform precision medicine strategies. Demographics of inflammatory bowel disease (IBD) patients in the entire SPARC IBD cohort are given in Supplementary Table 1. Within the SPARC IBD registry, clinical variables collected included symptom assessment (quality of stool as measured by Bristol stool scale, abdominal pain, stool frequency, rectal bleeding, urgency, and quality-of-life index). Stool samples were obtained within 3 to 6 months of colonoscopy—if stool samples were collected concurrent with colonoscopy, they were obtained as the first stool of the day prior to bowel preparation. Fecal calprotectin was obtained at time of stool sample collection. Fungal sequencing data were obtained through 18s Internal Transcribed Spacer (ITS2)–based deep sequencing of fungal ribosomal DNA from fecal samples. Tissue samples were obtained as up to 5 pinch biopsies from the cecum (or most proximal extent of the exam) and the rectosigmoid junction (at 20 cm). Further details of the cohort have been previously published.⁹

Inclusion/Exclusion Criteria for Study Cohort

Our study cohort was derived from within the SPARC IBD cohort. Data from 98 patients in the SPARC IBD registry were utilized if they had a history of UC, underwent colonoscopy, and had available ITS2 fungal mycobiome sequencing data (Supplementary Figure 1).

Assessment of Mycobiome

ITS sequence generation

ITS-based deep sequencing of fungal ribosomal DNA from fecal samples from the SPARC IBD registry was performed by Diversigen. DNA from 100 mg of fecal sample per subject was extracted using MagAttract PowerSoil DNA EP kit (Qiagen) using mechanical-based lysis with garnet beads. Samples were prepared using AccuPrime High Fidelity kit (Invitrogen) to amplify ITS2 (ITS3/ITS4). Sequencing was performed using the 2 × 300 bp paired-end protocol. The primers used for amplification contained adaptors for MiSeq sequencing (Illumina) and single-end barcodes allowing pooling and direct sequencing of polymerase chain reaction products. Target-specific primer sequences were: ITS3F: 5ʹ GCATCGATGAAGAACGCAGC 3ʹ and ITS4: 5ʹ TCCTCCGCTTATTGATATGC 3ʹ. Libraries were quantified with Quant-iT PicoGreen dsDNA assay (Invitrogen) and confirmed using gel electrophoresis. The median number of reads per sample was 15 030 (minimum = 1, maximum = 127 978).

ITS sequencing analysis

Sequences were quality filtered and denoised using Quantitative Insights Into Microbial Ecology (QIIME2) (version 2021.11).^10,11 Sequences were checked for quality and trimmed for barcodes and polymerase chain reaction primers, with low-quality reads below a Q-score of 30 dropped using DADA2 (version 2022.8.0).¹¹ Read-quality profiles for forward and reverse reads are shown in Supplementary Figure 2. Reads were binned for a minimal sequence length of 280 forward and 261 reverse. Results from quality filtering, denoising, merging, and chimera removal for each sample are given in Supplementary Table 2, with 82 samples remaining following sample filtering. We built fungal community matrices from the resulting unique amplicon sequence variants (ASVs). Sequence alignment was performed with MAFFT¹² and phylogenic trees were built with FastTree.¹³ Taxonomy was assigned with a fitted classifier using the UNITE (version 8.3) fungal reference data.^14,15

Diversity and abundance analysis

The R package Phyloseq (version 1.4; R Foundation for Statistical Computing) was used for evaluating diversity and abundance metrics, including Shannon index, UniFrac, and differential abundance at various taxonomic levels.¹⁶ We utilized observed ASVs and Shannon diversity indices to evaluate alpha diversity. We additionally calculated Hill numbers (Shannon effective numbers). For beta diversity, we used weighted UniFrac to generate a distance matrix, followed by nonmetric multidimensional scaling to ordinate the generated distances. To calculate proportions, we assessed the number of reads assigned to each phylum or genus divided by the total number of reads. To calculate fractional prevalence of phylum and genus, we assessed the number of patient samples with a specific phylum or genus, divided by the total number of patient samples.

To calculate the differential abundance of specific taxa across comparison groups in patients with UC, we utilized DESeq2 (version 3.15).¹⁷ DESeq2 employs a negative binomial generalized linear model to obtain maximum likelihood estimates for an ASV’s log fold change between 2 conditions (ie, patients with endoscopic activity vs endoscopic remission) and has increased sensitivity in smaller datasets.¹⁸ Log fold changes are then used with the Wald test for significance. Count tables were filtered to remove low-prevalence ASVs (<1% prevalence) to ensure that rare ASVs did not dominate comparisons. In a multivariate analysis, we adjusted for potential confounding effects of biologic therapy on fungal groups in categories of endoscopic inflammation vs remission. To perform the multivariate analysis, we utilized the multifactorial model included in DESeq2 to control for the covariates of interest (biologic exposure, age, and sex). Heat trees were generated using the metacoder package (version 0.3.5) in R¹⁹ to visualize relative abundance of taxa using counts across compared categories.¹⁹

Definitions

Endoscopic activity was defined as an MES > 0. Endoscopic remission was defined as a MES = 0. To define histologic activity, we evaluated digitized histology slides available through the SPARC IBD registry of biopsies obtained from their colonoscopy. Three independent pathologists blinded to clinical data scored each sample using the Nancy histologic index (NHI).²⁰ Histologic activity was defined as an NHI > 1, while histologic remission was defined as an NHI of 0 or 1. Endohistologic remission was defined as MES of 0 with an NHI of 0 or 1. Biologic exposure was defined as use of any thiopurines, anti-tumor necrosis factor (TNF) inhibitors, or vedolizumab prior to or up until the time of colonoscopy. Biologic-naïve patients were never exposed to thiopurines, anti-TNF therapy, vedolizumab, or any other biologic medications or small molecule therapies used in IBD and were otherwise untreated or on mesalamine therapy prior to or up until the time of colonoscopy.

Clinical assessment was based on clinical scores obtained from the SPARC IBD registry at time of colonoscopy and stool sample collection. Clinical scores were assessed for patients as follows: (1) Bristol stool scale (0-10); (2) abdominal pain (0 = none, 1 = mild, 2 = moderate, 3 = severe); (3) stool frequency score (0 = normal bowel movements, 1 = 1-2 stools over baseline, 2 = 3-4 stools over baseline, 3 = 5 or more stools); (4) rectal bleeding score (0 = no bleeding, 1 = blood in <50% of bowel movements, 2 = blood in >50% of bowel movements, 3 = passing blood alone); (5) urgency (0 = none; 1 = mild urgency—cannot wait 15 minutes; 2 = moderate urgency—I need to get to the bathroom in 2-5 minutes; 3 = severe urgency—I cannot wait 2 minutes); and (6) quality-of-life index as reported by the patient (0 = generally well, 1 = slightly under par, 2 = poor, 3 = very poor). Fecal calprotectin was obtained at time of colonoscopy and stool sample collection.

Outcomes

We described the alpha and beta diversity of the mycobiome and differential abundance of taxa among patients with UC with (1) endoscopic activity vs endoscopic remission, (2) endohistologic activity vs endohistologic remission, and (3) biologic exposure vs biologic-naïve.

Statistical Analysis

RStudio (version 2022.07.1) was used for all analyses and figure preparation. To test whether the observed number of ASVs, Shannon diversity index, or Shannon effective numbers differed significantly between comparators, we performed a Wilcoxon rank sum test (P < .05). To evaluate significance for beta diversity, we used permutational analysis of variance to calculate significance between groups (P < .05). For differential abundance analysis, we used adjusted P < .1 for discovery of altered taxa for DESeq2, with P values corrected for multiple testing using the Benjamini-Hochberg method.^21,22

Results

Sample Cohort

In this study, we examined clinical characteristics, endoscopic and histologic assessments, and 18s ITS2 fungal mycobiome data among a cohort of 98 UC patients (Supplementary Figure 1). Mean cohort age was 44.9 ± 14.3 years, with 52 women and 46 men. Median disease duration was 2 (interquartile range 0-4) years. During their duration of disease activity, 31.6% were biologic-naïve, while 67.3% of patients were exposed to anti-TNFs, thiopurines, or vedolizumab. Patients reported variable clinical symptoms at time of colonoscopy, as assessed by the Bristol stool scale, abdominal pain, stool frequency, rectal bleeding, and quality of life (Table 1).

Table 1.

Study cohort characteristics.

	Total Cohort	Endoscopic Status			Endohistologic Status			Biologic Use
		Activity	Remission	P	Activity	Remission	P	Exposed	Naive	P
n	98	25	28	—	19	23	—	66	31	—
Age, y	44.9 ± 14.3	48.5 ± 13.4	47.7 ± 15.7	.76	46.2 ± 14.5	45.1 ± 12.1	.99	48.6 ± 13.8	51.6 ± 14.9±	.37
Female, %	53	52	64.3	.36	68.4	60.9	.61	50	61.3
Disease duration, y	2 (0-4)	0 (0-4)	2 (0-4)	.44	0 (0-1.75)	2 (0-4)	.27	2.65 (0.25-4)	3 (0-4.25)	.96
Fecal calprotectin, mg/g	34.9 ± 84.7	99.3 ± 143.3	8.78 ± 14.7	<.05a	89.7 ± 152.1	23.6 ± 49.1	.09	39.3 ± 90.3	40.93 ± 77.1	.26
Bristol stool scale	4 (4-5)	4 (4-6)	4 (4-4)	.14	5 (4-6)	4 (4-4.5)	.20	4 (4-5)	4 (4-5.75)	.46
Abdominal pain score	0 (0-1)	0.5 (0-1.25)	0 (0-0)	<.05a	0 (0-1.5)	0 (0-0)	<.05a	0 (0-1)	0 (0-1)	.92
Stool frequency score	0 (0-1)	1 (0-2)	0 (0-1)	.10	1 (0-1)	0 (0-0)	<.05a	0 (0-1)	0 (0-1)	.88
Rectal bleeding score	0 (0-0)	0 (0-1)	0 (0-0)	.08	0 (0-0)	0 (0-0)	.54	0 (0-0)	0 (0-0)	.34
Urgency score	1 (0-1)	1 (0-2)	1 (0-1)	.09	1 (0-2)	1 (0-1)	.13	1 (0-1)	1 (0-2)	.16
Quality-of-life score	0 (0-1)	1 (0-1)	0 (0-0)	<.05a	1 (0-1)	0 (0-0)	<.005a	0 (0-1)	0 (0-1)	.90

Values are mean ± SD or median (interquartile range), unless otherwise indicated.

^a P value < .05.

Within this cohort, all patients underwent colonoscopy (n = 98) (Supplementary Figure 1). However, because we were interested in evaluating associations between fungal populations and endoscopic/histologic status, we only assessed patients who underwent colonoscopy within 6 months of stool sample collection. A total of 53 patients had colonoscopy and stool sample collection within 6 months of each other (median interval 4 [interquartile range 1-5] months).

Among patients with available endoscopic-mycobiome data (n = 53), 47% had endoscopic activity (MES > 0), while 53% of patients had complete endoscopic remission (MES 0). Of those with endoscopic activity, 72% had MES = 1 and 28% had MES ≥ 2. Compared with endoscopic remission (MES = 0), patients with endoscopic activity (MES > 0) reported significantly increased abdominal pain (P < .05), with numerically increased stool frequency, urgency, and reduced quality of life. Additionally, compared with endoscopic remission, patients with endoscopic activity had significantly increased fecal calprotectin (99.3 mg/g vs 8.78 mg/g; P < .05) (Table 1).

Among the 53 patients with endoscopic-mycobiome data, 42 patients also had concurrent histologic slides available, which we retrospectively assessed and independently validated using the NHI. Among patients with available endohistologic-mycobiome data (n = 42), 40.5% had active histologic disease, while 59.5% had histologic remission. Furthermore, using concomitantly available endoscopic scores, we determined that 45.2% of patients had active endohistologic disease, while 54.8% were in complete endohistologic remission (Table 1). Compared with endohistologic remission, patients with endohistologic activity reported significantly increased abdominal pain (P < .05) and stool frequency (P < .05). Patients with endohistologic activity also had an increased fecal calprotectin compared with endohistologic remission (89.7 mg/g vs 23.6 mg/g), although that difference was not statistically significant (P = .09) (Table 1).

Among patients with biologic exposure vs those who were biologic-naïve, we saw no differences in demographics, clinical assessments, and fecal calprotectin (Table 1).

Overview of Fecal Fungal Diversity in UC

Among fungal sequences analyzed from 98 patients, following quality filtering, denoising, merging, and chimera removal for each sample, 82 samples remained (see Methods). We examined the phylogeny and abundance of fungi in fecal samples and identified a diverse fungal community of 79 unique fungal ASVs across 82 patients (Figure 1). Of all reads, 73% belonged to phylum Ascomycota, 11.1% belonged to Basidiomycota, 0.15% belonged to Mucoromycota, and 15.8% were unidentified. Within Ascomycota, 83.6% of reads mapped to the order Saccharomycetales, containing genus Candida, Pichia, and Saccharomyces. Within Basidiomycota, the order Tremellales made up 16.3% of reads, while Malasseziales composed 1.8% of reads. We additionally examined the fractional prevalence of phylum and genera across all 82 patients (Supplementary Figures 3A, B). We found that Ascomycota was present in 73.1% of fecal samples, Basidiomycota in 8.5%, and Mucoromycota in < 1%, with an unclassifiable fungal phylum present in 43.9% of patients. Among genera, we found that the genus Blumeria was found in 28% of patients, Candida in 32%, Saccharomyces in 43.9%, and Penicillium in 17%, with 51% of patients having an unclassifiable fungal genus. To ensure that identified and unidentified ASVs were not systematically different, we compared principal coordinate analysis plots using weighted UniFrac distances between both groups and saw similar distributions (Supplementary Figure 4).

Figure 1.

Fungal taxa abundance in ulcerative colitis. Abundance heat tree depicts the abundance of each identified amplicon sequence variant (ASV) across all patient samples (n = 82). Nodes represent each taxonomic rank from kingdom (fungi, center) to genus (tips of each branch). The size of the node corresponds to the number of taxa in that group, and the color intensity corresponds to mean abundance of that taxa across patient samples (dark green samples are highly abundant, and light yellow and light gray represent low abundance).

Fungal Diversity in Endoscopic Activity

We next compared the diversity of the mycobiome between patients with UC in endoscopic remission (n = 25) and those with endoscopic activity (n = 28). We saw no significant differences in alpha diversity using the observed number (10.65 vs 9.21; P = .21), Shannon diversity indices (1.06 vs 1.03; P = .93) (Figure 2A, Supplementary Table 3), or Shannon effective numbers (3.67 vs 3.75; P = .93). We then identified clusters of similar fungal microbiota by calculating pairwise distances between each sample. We found that endoscopic activity had no detectable effect on fungal clustering, with similar beta fungal diversity among patients with endoscopic activity vs remission (P = .397; PERMANOVA) (Figure 2B, Supplementary Table 4).

Figure 2.

Fungal diversity in endoscopic activity vs endoscopic remission. A, Alpha diversity: amplicon sequence variants observed and Shannon diversity indices with boxplots representing samples colored by endoscopic status. B, Beta diversity: nonmetric multidimensional scaling plot of weighted UniFrac distance with each sample colored by endoscopic status.

We examined the differences in phylum and genus between patients with endoscopic activity vs endoscopic remission. With endoscopic remission, 58.5% of reads belonged to Ascomycota, including the genera Saccharomyces (8.2%) and Candida (5.6%). Notably, an unidentified group of fungi made up 35.5% of reads in endoscopic remission, with Basidiomycota (5.1%) and Mucoromycota (0.9%) composing the remaining reads. In endoscopic activity, 87.1% of reads belonged to Ascomycota, including genera Saccharomyces (53.3%) and Candida (24.3%). The remaining reads were attributed to unidentified fungi (7.0%), Basidiomycota (5.9%), and Mucoromycota (<1%).

Differential Abundance in Endoscopic Activity

We examined differential abundance of fungal taxa within patients with endoscopic activity vs endoscopic remission using a negative binomial model (DESeq2). We found that during endoscopic activity, Saccharomyces was increased (log2 fold change = 4.54; adjusted P < 5 × 10^-5) as was Candida (log2 fold change = 2.56; adjusted P < .03) (Figure 3A, Table 2). We found that in endoscopic remission, Penicillium was increased (log2 fold change = 4.94; adjusted P < 5 × 10^-7). We generated a heat tree depicting the log mean ratio of the relative abundance of taxa during endoscopic activity vs endoscopic remission (Figure 4).

Figure 3.

Differentially abundant taxa across the endohistologic spectrum in ulcerative colitis. Log2 fold changes of differentially abundant amplicon sequence variants at the genus level were determined using negative binomial generalized linear modeling (DESeq2) among patients in (A) endoscopic activity vs remission, (B) endohistologic activity vs endohistologic remission, and (C) endoscopic activity after adjusting for age, sex, and biologic exposure.

Table 2.

Altered taxa in endoscopic activity vs endoscopic remission.

Base Mean	Log2 Fold Change	SE	P Value	Adjusted P	Phylum	Family	Genus
5810.68	4.54	1.02	8.85 × 10-6	4.87 × 10-5	Ascomycota	Saccharomycetaceae	Saccharomyces
62.67	2.56	0.95	7.00 × 10-3	.03	Ascomycota	Saccharomycetales incertae sedis	Candida
148.89	-4.94	0.87	1.30 × 10-8	1.43 × 10-7	Ascomycota	Aspergillaceae	Penicillium

Adjusted P value for significance < .05.

Figure 4.

Relative abundance of fungal taxa in endoscopic activity and remission. Log mean ratio of relative abundance of taxa in endoscopic activity and remission. Nodes represent each taxonomic rank from kingdom (fungi, center) to genus (tips of each branch). The size of the node corresponds to the number of taxa in that group, with increasing color intensity representing taxa that are highly abundant during endoscopic activity (dark blue) and low color intensity representing taxa that are highly abundant during endoscopic remission (light blue). ASV, amplicon sequence variant.

Fungal Diversity and Abundance in Endohistologic activity

We next performed diversity and differential abundance analyses among patients with ongoing endohistologic activity (n = 19) vs endohistologic remission (n = 23). The mean number of observed taxa were similar (11.38 vs 5.5; P = .36), as were Shannon diversity indices (1.05 vs 1.09; P = .95) (Supplementary Figure 5A, Supplementary Table 5) and Shannon effective numbers (3.75 vs 3.91; P = .95). Beta diversity showed similarities between endohistologic activity vs remission (P = .541) (Supplementary Figure 5B, Supplementary Table 6). In an exploratory analysis, we examined differential abundance testing of phylum and genera between patients with endohistologic activity and remission. We found that Saccharomyces had an increase in endohistologic activity (log2 fold change = 2.78; adjusted P < .08). We found that in endohistologic remission, there was an increase in Penicillium (log2 fold change = 5.45; adjusted P < 5 × 10^-10) (Figure 3B, Supplementary Table 7).

Fungal Diversity and Abundance in Biologic Exposure

We then performed diversity and differential abundance analyses among patients who were exposed to biologic medications (n = 66) (either anti-TNF or anti-integrin therapy) or who were naïve to biologic medications (n = 31). The mean number of observed taxa was similar (10 vs 10.1; P = .13), as were Shannon diversity indices (1.18 vs 0.96; P = .25) and Shannon effective numbers (4.29 vs 3.24; P = .25) among patients who were biologic exposed vs biologic-naïve (Supplementary Figure 5C, Supplementary Table 8). We found no difference in beta diversity between patients who were biologic-naïve vs biologic exposed (P = .30) (Supplementary Figure 5D, Supplementary Table 9). We then examined differential abundance testing of phylum and genera between patients with biologic exposure vs patients who were biologic-naïve. We found no significant differences in fungal genera between patients exposed to biologics vs patients who were biologic-naïve.

Multivariate Model of Differential Abundance in Endoscopic Activity and Biologic Exposure

We evaluated the differential abundance of genera among patients with UC during endoscopic activity vs endoscopic remission after controlling for the effects of age, sex, and biologic exposure. Adjusting for these covariates, we found that UC patients with endoscopic activity continued to have significant increases in Saccharomyces (log2 fold change = 7.76; adjusted P < 1 × 10^-15) and Candida (log2 fold change = 7.28; adjusted P < 1 × 10^-8) compared with endoscopic remission, with changes also observed in several other taxa, including an unidentified genus (Figure 3C, Supplementary Table 10).

Discussion

In this secondary analysis of a multicenter, longitudinal, prospective cohort of patients with UC, we evaluated changes in the fecal fungal community across the spectrum of endoscopic activity, histologic activity, and exposure to biologic medications. Overall, we report here that among patients with UC, Ascomycota is enriched in endoscopic activity, with expansion of Saccharomyces and Candida compared with endoscopic quiescence.

Both Saccharomyces and Candida are highly abundant genera in the human gut. Saccharomyces has been previously linked to inflammation.²³ Increases in the order Saccharomycetales, containing Saccharomyces, have been previously observed on inflamed colonic mucosa in UC.⁷ ASCA antibodies are recovered from the serum of 10% to 15% of UC patients.^2,3 Mouse models lavaged with S. cerevisiae developed worsened colitis and increased gut permeability.²⁴ Intriguingly, in contrast to our findings, a prior study revealed that Saccharomyces was reduced in UC during a flare compared with remission, although that study did not use endoscopic parameters to define activity.⁶ Differences in diet may explain cohort-related discrepancies, as this study was performed in a single center in Paris, France,⁶ compared with our study utilizing patients throughout the United States. Given that we found that Saccharomyces was increased on multivariate analysis lends further credence to our observation that this taxa in enriched during inflammation in UC.

Candida is also a highly abundant, commonly detected fungi in the gut, has been observed to drive murine colitis, and is elevated in several IBD cohorts.^6,23,25 Gut Candida may induce various antifungal antibodies, acting as an immunogen for ASCA.^26,27 Most recently, the abundance of Candida prognosticated responses to FMT as a therapy in UC.⁸ Specifically, patients with high Candida prior to FMT appear to have the best response to stool donation, with reduced abundance of Candida post-FMT suggestive of ameliorated disease severity. Such results are consistent with the observations in our cohort, in which high Candida was linked to endoscopic severity, with low Candida linked to quiescence. Our results are consistent with a second study that reveals increased Candida in the mucosal mycobiome of patients with UC compared with healthy individuals.²⁸

First, given that immunosuppressives drugs have been previously linked to an increased expansion of Candida,²⁹ it was possible that increases in Candida during endoscopic activity might be related to increased use of immunosuppressive medications. However, we did not find that biologic exposure was associated with increased Candida. Second, when we adjusted for biologic exposure, Candida remained enriched in patients with endoscopic activity, suggesting this effect is independent of immunosuppressive exposure.

Overall, these findings are intriguing in the context of prior studies that reported that UC polymorphisms in Dectin-1, a fungal sensor in the human gut, strongly induce a T helper 17 cell response; the increase of Saccharomyces and Candida, both high-abundance organisms, during inflammation may reflect enhanced T helper 17 cell responses through such fungal sensors.

We did not find differences in fungal diversity between patients with endoscopic activity and remission. While bacterial diversity has been previously shown to be reduced during disease activity in UC,^30,31 inflammatory status does not appear to associate with fungal diversity in IBD. We also compared the abundance of fungi among patients with endohistologic activity against complete endohistologic quiescence. Ongoing histologic activity—despite endoscopic remission—may provide an altered environment for the growth of specific fungal groups. Similar to our findings comparing endoscopic activity and remission, we found that Saccharomyces was significantly increased in endohistologic activity.

Despite using the most updated version of a fungal reference database,¹⁴ 17.2% of reads in our cohort were still assigned to unidentified fungi. Prior cohorts have also reported similar proportions of reads assigned to unidentified fungi in IBD.^6,28 We also did not evaluate abundance at the species level, due to the difficulty of assigning reads at such depth. In the future, more comprehensive reference fungal genomes gathered from varying populations may help resolve the problem of unidentified taxa.³²

Additionally, our analyses were compositional; therefore, our report that Saccharomyces and Candida are increased in active UC actually reflects that the fractions of these genera increased relative to other genera. Future studies utilizing quantitative profiling may provide additional insight into absolute changes in fungal abundances and disease phenotype. We also did not report on bacterial communities in this study. Prior studies in IBD report that fungal communities are linked to changes in bacteria that may play critical roles in driving or suppressing inflammation in the gut.^6,8 Future studies examining fungal-bacterial dynamics during inflammation vs remission are warranted to explore the role of interkingdom relationships in UC pathogenesis. Dietary information was also not explored in this study, and it is possible that some of the genera identified are dietary contaminants. Studies exploring fungal-bacterial relationships longitudinally and serially in UC, with concomitant measurement of diet, would be necessary to further explore these relationships.

The strengths of this study include that we utilized a well-characterized, prospective, longitudinal cohort enrolling patients from 13 centers across the United States. Utilizing this cohort allowed for consistency in data and biosample collection methods, with multiple investigators involved in reviewing extracted data variables.⁹ We also only utilized ITS2 data from stool samples obtained in close temporal proximity to colonoscopy. We further established validity of our endohistologic classifications by examining clinical symptoms and concurrently obtained fecal calprotectin in these subgroups. Novel in this study, we evaluated the fungal mycobiome among various inflammatory categories of UC, defined by validated endohistologic criteria, and accounted for changes potentially induced by age, sex, and biologic exposure. Limitations include the relatively small sample size of the population studied, a lower number of severely active UC patients, a single time point of stool sample collection, and a proportion of unidentified fungi in the reference fungal database.

Overall, we report that among patients with UC, endoscopic inflammation is linked with increased Saccharomyces and Candida, even after adjusting for age, sex, and exposure to biologic therapies. Our results are consistent with prior observations that Saccharomyces and Candida may be associated with increased inflammation in IBD. In the light of recent study supporting that FMT may be more useful in specific fungal mycobiomes to treat UC, future studies exploring the fungal mycobiome in relation to bacterial constituents longitudinally across endoscopic and histologic parameters may provide further insight into predicting outcomes and personalizing therapeutic approaches.

Author Contribution

Conception: K.H., N.Z., H.C., S.S., D.M., S.J. Data abstraction: K.H., K.A., L.M.K., M.F., H.C., S.J. Data analysis and interpretation: K.H., N.Z., M.C., H.C., S.S., D.M., S.J. Drafting of manuscript: K.H., N.Z., S.S., D.M., S.J. Critical review of manuscript: K.H., N.Z., S.S., D.M., S.J. Statistical analysis: S.J. Guarantor of article: S.J. The manuscript, including related data, figures and tables has not been previously published and that the manuscript is not under consideration elsewhere.

Funding

The project described was supported by the National Center for Advancing Translational Sciences of the National Institutes of Health, Award Number KL2TR002545. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. Additionally, this work was supported by the Charlton Research Award and Natalie V. Zucker Research Center for Women Scholars Award, both from the Tufts University School of Medicine.

Conflicts of Interest

S.S. has received institutional research grants from AbbVie and Pfizer and personal fees from Pfizer. All other authors disclose no conflicts.

Data Availability

ITS2 sequencing, clinical data, metagenomic data, and tissue pathology were obtained from the Crohn’s & Colitis Foundation. For access to ITS2 and clinical metadata, researchers may contact the Crohn’s & Colitis Foundation and obtain data through IBD Plexus (https://www.crohnscolitisfoundation.org/research/grants-fellowships/ibd-plexus).